[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] ProcD issues



On 4/24/2013 4:26 PM, John Lambert wrote:
We've been having some trouble with ProcD. Whenever a machine changes
states (usually going from backfill to unclaimed) ProcD has the
following error on Windows 8 machines (not entirely sure if it's
exclusive to them...)

04/23/13 17:39:41 CreateFile error: 2
04/23/13 17:39:41 ProcFamilyClient: failed to start connection with ProcD
04/23/13 17:39:41 unregister_subfamily: ProcD communication error
04/23/13 17:39:41 ERROR "ProcD has failed" at line 621 in file
c:\condor\execute\dir_1756\userdir\src\condor_utils\proc_family_proxy.cpp

The ProcD.STARTD logs don't mention any errors, so I'm not sure what
happened.

Anyway, I was wondering what alternatives to ProcD existed. The ProcD
wisdom page mentions a killfamily class method. Is this a UNIX exclusive
thing? How can I configure the rest of my daemons to not use ProcD? Any
light you guys can shine on this would be greatly appreciated.


Hi John -

Thanks for the heads-up re potential issues w/ the PROCD on Windows 8. We do not yet do automated regression testing on Windows 8 but will be starting soon (our build/test pool has Windows 7, and we just added in a Windows 8 machine but haven't fully configured it yet....soon...).

To disable the procd you can set
  USE_PROCD = False
in your condor_config and then restart HTCondor (I don't think a reconfig will do it here, I think you'll need a full stop and start of the service). This setting works the same on all platforms. In general use of a PROCD is advisable for scalability issues (see below) esp if you have an execute machine with a lot of slots, although the scalability impact of disabling the procd may actually be less on Windows than on Unix-type platforms. But everything should still work with the PROCD disabled. Info on USE_PROCD knob from the manual cut-n-pasted below.

p.s. look forward to meeting you next week in Madison @ HTCondor Week 2013!

regards,
Todd


USE_PROCD
This boolean variable determines whether the condor_procd will be used for managing process families. If the condor_procd is not used, each daemon will run the process family tracking logic on its own. Use of the condor_procd results in improved scalability because only one instance of this logic is required. The condor_procd is required when using privilege separation (see Section 3.6.14) or group ID-based process tracking (see Section 3.12.11). In either of these cases, the USE_PROCD setting will be ignored and a condor_procd will always be used. By default, the condor_master will start a condor_procd that all other daemons that need process family tracking will use. A daemon that uses the condor_procd will start a condor_procd for use by itself and all of its child daemons.